Few sweeps - long sweep time or many sweeps - short time?

[solutionseeker]

Would it be better to do fewer sweeps but with slower sweep speeds? I'm wondering if it is better to do lots of fast sweeps, or a few slower ones. What do you think?

[Lüder]

Dear solutionseeker,
do you have any Idea about your Signal?
I case of wireless broadcast I have short packages of emission and so I perform a lot fast scans. And you will see if the resulting plots get smooth you have enough data.
But for conducted emission I just do three slow scans and got everything, that's is faster then a lot fast scans.

So depending on your signals it is different.

I got a stupide setup with a lot slow scans as well, for not experienced colleagues or run it during lunch time so I am sure nothing is missing.

Best regads

[joro]

It not easy to give a single answer on this question that is working in all situations.

There are a few possibilities to perform an emission measurement of a frequency range:

1. Use a (scanning) receiver that measures on each frequency for a fixed amount of time (for example: 10 ms).
2. Use a spectrum analyzer that sweeps with a fixed sweep time (for exampe: 200 ms) over the complete frequency range
3. Use an FFT analyzer to measure the frequency range

There are many differences between and measurement principles between these measurement devices. But if we go into detail on the timing differences, we can describe.
Let's assume that a measurement range is measured from 80 MHz - 1000 MHz, using a 120 kHz RBW, with a step of 60 kHz.

1. Every frequency (which is (1000 MHz - 80 MHz) / 0.06 MHz = +-15335 frequencies) is measured for 10 ms. The total measurement for a single scan thus takes a little bit more than 153 seconds. Due to this long measurement time, very often only a single scan of the frequency range is performed. Each individual frequency is measured for 10 ms.
2. With the spectrum analyzer the sweep of the frequency range takes 200 ms. However, every individual frequency is measured for a indefinite short time, because the sweeping analyzer is continuously moving, and never stops at a single frequency to measure a specific single frequency. Because of the short sweeptime, and to prevent that every frequency is only measured very short, a lot of sweeps are normally taken. This can be 10 sweeps (with increases the total measurement to 2 seconds), 20 sweeps (total: 4 seconds) or even 100 sweeps (total: 20 seconds).
3. The FFT analyzer very often measures a frequency down-mixed specific bandwidth (for example: 60 MHz) with a fast AD-Convertor (for example: 50 ms) and uses the FFT algorithm to determine the frequencies and their amplitudes int that block. All the frequencies in this 60 MHz bandwidth are thus measured for 50 ms, but to perform a measurement for the total frequency range, 16 blocks ((1000 MHz - 80 MHz) / 60 MHz) of 60 MHz thus have to be measured which takes a total time of 800 ms.

Measuring a CW signal that always has a fixed amplitude is not a problem for all these 3 combinations, every measurement configuration will correctly determine the amplitude of a CW signal. The problem is however raised when a in amplitude fluctuating signal is measured. The signal repetition rate (also very often called: pulse repetition rate), which specifies the time interval that the signal reaches the maximum amplitude again, has the most influence, in combination with the Duty Cycle of the signal.

Assume that the EUT emits a signal on a specific frequency that is only present for 100 ms, is not present for 900 ms. The signal repetition rate is thus 1 second, and the duty cycle is 10%.

1. If the (scanning) receiver is measuring the specific frequency in the time period that the signal is not present, it will not measure and not detect the signal. To correctly determine and measure the signal a measure time of 1 second should be used. The total measurement of the complete frequency range than will become roughly 15300 seconds (> 4 hours).
2. When the spectrum analyzer is measuring the complete frequency range, there is a 50% chance (100 ms time in a 200 ms sweeptime) that the signal is measured when the signal is transmitting. The next 4 sweeps the signal is not transmitting, and will thus not be measured, the 6th sweep there is again a 50% chance that the signal is measured. Performing 10 sweeps is probably not enough to measure the signal, but the chance that the signal will be measured when 30 sweeps are taken is very high (however not guaranteed). The total measurement time for the frequency range will then be 6 seconds.
3. When the FFT analyzer is measuring the complete frequency range, there is a high chance that another 60 MHz block is being measured when the signal is transmitting. Two FFT blocks can be measured during the period that the signal is transmitting. Thus the chance is 12.5% that the signal is measured. For the same reason as the spectrum analyzer, thus multiple sweeps are being taken to have a higher chance of measure the signal when it is transmitting. For example 20 sweeps is probably sufficient. The total measurement time will then be 16 seconds.

This is just a single example. For other signals, with another signal repetition rate and duty cycle, the correct configuration of sweeptime and amount of sweeps is probably different. This also indicates that it very often is important to know on before hand the type of signals, and their signal repetition rate, that an EUT is transmitting. We also know EUT's that only transmit a signal for a few milli-seconds every few seconds. And there are EUT's that only have a operating mode for some seconds, every single minute. Each of this signals do require some knowledge of the signal, and the correct configuration of the measurement device to be able to correctly measure the signals.

The original question is: should one use 'few sweeps - long sweep time' or 'many sweeps - short sweep time'? In general one can conclude that more measurement data gives more information, and that many sweeps with a shorter sweep time over the complete frequency range, gives a higher chance that signals with a long signal repetition time are detected. It is however important that the amount of sweeps and the total measurement time is then sufficiently enough longer than the signal repetition rate to have a high chance of measuring the signal. The most important conclusion is that it will have benefit to know in advance the type of signals and their characteristics that are emitted by the EUT in advance.

I hope that this gives you some idea on how to make a decision between 'few sweeps - long sweep times' and 'many sweeps - short time'. Please let us know your opinion and experience or suggestions.

[joro]

To emphasize the benefit of knowing the type of signal in advance, I would also like to give the following example.

When one knows the type of signal that is emitted by an EUT, it is of course also possible to use this knowledge to configure the emission test.
Assuming that the signal of the previous post (1 second repetition time, 10% Duty Cycle) is known in advance, and is transmitted at 860 MHz.

A Radiated Emission test can then be configured with the Spectrum Analyzer or the FFT analyzer that only measures 859 MHz to 861 MHz with a sweeptime of 25 ms.
When this test is configured to take 40 sweeps (so the total frequency range measure time becomes 1 second), it is guaranteed that the that signal is measured correctly, in an emission test that only takes a little bit more than 1 second.
In a Multiband Radiated Emission test, this small frequency range can of course also be added as a separate band, in addition to the original wide frequency range band (80 MHz - 1000 MHz).